Вы находитесь на странице: 1из 15

Tectonophysics 417 (2006) 269 – 283

www.elsevier.com/locate/tecto

3-D velocity structure of the 2003 Bam earthquake area (SE Iran):
Existence of a low-Poisson's ratio layer and its
relation to heavy damage
Hossein Sadeghi a,⁎, S.M. Fatemi Aghda b,1 , Sadaomi Suzuki c,2 , Takeshi Nakamura d,3
a
Earthquake Research Center, also at Department of Geology, Ferdowsi University of Mashhad, Mashhad 91775-1436, Iran
b
Department of Geology, Tarbiat Moallem University, Tehran 15614, also at Natural Disaster Research Institute, Tehran 19395-4676, Iran
c
Tono Research Institute of Earthquake Science, Mizunami 509-6132, Japan
d
Department of Earth and Planetary Sciences, Faculty of Sciences, Kyushu University, Fukuoka 812-8581, Japan
Received 1 July 2005; received in revised form 2 January 2006; accepted 31 January 2006
Available online 15 March 2006

Abstract

To understand the generation mechanism of the Bam earthquake (Mw 6.6), we studied three-dimensional VP, VS and Poisson's
ratio (σ) structures in the Bam area by using the seismic tomography method. We inverted accurate arrival times of 19490 P waves
and 19015 S waves from 2396 aftershocks recorded by a temporal high-sensitivity seismic network. The 3-D velocity structure of
the seismogenic region was well resolved to a depth of 14 km with significant velocity variations of up to 5%. The general pattern
of aftershock distribution was relocated by using the 3-D structure to delineate a source fault for a length of approximately 20 km
along a line 4.5km west of the known geological Bam fault; this source fault dips steeply westward and strikes a nearly north–
south line. The main shallow cluster of aftershocks south of the city of Bam is distributed just under the minor surface ruptures in
the desert. The 3-D velocity structure shows a thick layer of high VS and low σ (minimum: 0.20) at a depth range of 2–6 km. The
deeper layer, with a thickness of about 2 km, appears to have a low VS and high σ (maximum: 0.28) from 6 km depth beneath Bam
to a depth of 9 km south of the city. The inferred increase of Poisson's ratio from 2 to 10 km in depth may be associated with a
change from rigid and SiO2-rich rock to more mafic rock, including the probable existence of fluids. The main seismic gap of
aftershock distribution at the depth range of 2 to 7km coincides well with the large slip zone in the shallow thick layer of high VS
and low σ. The large slip propagating mainly in the shallow rigid layer may be one of the main reasons why the Bam area suffered
heavy damage.
© 2006 Elsevier B.V. All rights reserved.

Keywords: Bam earthquake; Aftershocks; Seismic velocity structure; Poisson's ratio; Shallow rigid layer; Arg-e-Bam fault

1. Introduction
⁎ Corresponding author. Fax: +98 511 842 1234.
E-mail addresses: sadeghi@seismo.um.ac.ir (H. Sadeghi), On December 26, 2003, a powerful earthquake in
fatemi@saba.tmu.ac.ir (S.M.F. Aghda), suzuki@tries.jp (S. Suzuki),
southeastern Iran caused catastrophic damage to the
nakamura@geo.kyushu-u.ac.jp (T. Nakamura).
1
Fax: +98 21 200 9453. ancient city of Bam, located on the Silk Road, and
2
Fax: +81 572 67 3108. neighboring villages with a collective population of
3
Fax: +81 92 642 2684. about 142,000. In terms of loss of life and casualties, this
0040-1951/$ - see front matter © 2006 Elsevier B.V. All rights reserved.
doi:10.1016/j.tecto.2006.01.005
270 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

earthquake was the worst to occur in that year anywhere Talebian et al. (2004), Binet and Bollinger (2005), and
in the world. The death toll was 26,371 people, nearly Fielding et al. (2005) indicated that the main rupture
19% of the population (Statistic Center of Iran, 2004) reached the surface some 5 km west of the Bam fault.
and tens of thousands were injured. An important Wang et al. (2004) used differential radar interferometry
cultural loss was the almost total destruction of the well- (D-inSAR) to determine the source parameters. They
known historic citadel Arg-e-Bam. This monument, suggested that the Bam earthquake ruptured a hidden or
declared by UNESCO as a World Heritage Site, is the new fault of about 24 km from (29.178N, 58.382E) to
biggest mud-brick structural complex in the world. Arg- (28.971N, 58.357E) that had an unusually strong
e-Bam is thought to be over 2000 years old. Since the asperity.
structure was well intact before the 2003 Bam The main shock was followed by a series of
earthquake, this earthquake is believed to be the largest aftershock events (e.g., Nakamura et al., 2005; Tatar et
to occur in this area in more than 2000 years. al., 2005). Nakamura et al. (2005) recorded several
Aside from the destruction of Arg-e-Bam, the main thousand aftershocks during the period February 6 to
reason for such massive damage to Bam may be the poor March 7, 2003. They located accurate hypocenters of
construction of the unreinforced mud-brick houses. 2789 aftershocks by the use of a 1-D velocity model.
Even so, the damage was disproportionately and The main distribution of aftershocks did not correspond
unexpectedly large given the magnitude of the earth- to the geological Bam fault. The epicenter distribution is
quake. Maximum acceleration of 0.98g was recorded in linearly more than 20km in parallel with a line about
the vertical component at the Bam accelerograph station 3.5 km west of the Bam fault. The hypocenter
in the center of Bam city by the Building and Housing distribution shows a nearly vertical trend or a slight
Research Center of Iran (BHRC; http://www.bhrc.gov. tendency to lie farther west with its depth increasing
ir/). The earthquake information initially provided by from 0 to 16 km. They proposed the name “Arg-e-Bam
the US Geological Survey, USGS (http://earthquake. fault” as the source fault to distinguish it from the Bam
usgs.gov/), was as follows: origin time 26/12/2003 at fault.
01:56:52 (UTC) and 05:26:52 (local time); epicenter Since the early 1980s, local earthquake tomography
29.004N, 58.337E; depth 10 km, and moment magni- has been successfully used to image lateral heterogene-
tude 6.6. Suzuki et al. (2004) estimated the start point of ities of the crust in seismogenic fault zones (see
the large slip in the source fault by using the aftershock Eberhart-Phillips, 1993 and references therein). Material
distribution and the S–P time recorded by the Bam properties in the earthquake source area would certainly
accelerograph station. Teleseismic focal mechanisms have influenced the initiation, propagation and termi-
from several groups (e.g., USGS; Yamanaka, 2003) nation of the earthquake rupture. The velocity variations
indicated a steeply dipping, right lateral strike-slip in the upper crust allow us to define the structure and
movement on a fault with a N–S trend. This agrees well geometry of faults at depth and to identify the structures
with the known tectonic pattern of the region (e.g., of seismogenic zones, especially in cases of hidden and
Walker and Jackson, 2002). The Bam fault, which has buried faults (e.g., Eberhart-Phillips, 1990; Chiarabba
long been recognized (e.g., Berberian, 1976; Hessami et and Selvaggi, 1997; Chiarabba et al., 1997, Zhao et al.,
al., 2003) for its distinctive flexure scarp, extends along 2004). In the present study we have applied seismic
the west side of the village of Baravat about 5km tomography to arrival time data of the Bam earthquake
southeast of Bam city. Just after the earthquake, it was aftershocks recorded by a temporal high-sensitivity
assumed that the main shock had occurred in the seismic network (Suzuki et al., 2004; Nakamura et al.,
geological Bam fault (e.g., Ahmadizadeh and Shakib, 2005). Arrival times of high-quality P waves and S
2004). However, nobody could find any clear evidence waves were collected. These data allowed us to
of dislocation on this existing Quaternary fault. determine detailed three-dimensional (3-D) VP and VS
Comparing the Bam earthquake and the 2000 Tottori structures in the source area of the Bam earthquake. We
earthquake in southwest Japan, Miyake et al. (2004) used the data to try to deduce variations in the physical
proposed that the main shock ruptured a shallow properties of rocks. Because rocks with differing
asperity on a fresh fault rather than on the Bam fault. physical states can have similar seismic velocities,
Waveform inversions using teleseismic data (e.g., seismic velocity alone is not a sensitive indicator of
Yamanaka, 2003; Yagi, 2003) have suggested the variable rock properties. For this reason, it is often
existence of a shallow asperity, i.e., a large slip area, useful to consider the ratios and products of seismic
for the Bam earthquake. Analyzing Envisat SAR parameters to differentiate 3-D variations in the
interferometry data before and after the earthquake, subsurface (Salah and Zhao, 2003). Poisson's ratio (or
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 271

VP / VS) is a key parameter in studying petrologic Makran belt. The Makran belt is the emerged portion of
properties of crustal rocks (Christensen, 1996) and can an accretionary prism resulting from the subduction of
provide tighter constraints on the crustal composition the Oman Gulf oceanic lithosphere (which forms part of
than can VP and VS alone. Poisson's ratio has proved to the Arabian plate) beneath the Iranian plate (Byrne et al.,
be very effective for clarification of the seismogenic 1992; McCall, 1997; Kopp et al., 2000). Earthquakes
behavior of crust, especially the roles of crustal fluids in have occurred mainly within the Zagros as a wide
the nucleation and growth of earthquake rupture (Zhao seismic belt (see USGS National Earthquake Informa-
et al., 1996, 2002, 2004). tion Center catalog, available at http://neic.usgs.gov/
neis/epic/epic.html). The high seismicity in Zagros
2. Tectonic setting might be due to the presence of thick layers of late
Precambrian to early Cambrian salt deposits allowing
The tectonics of the region (Fig. 1) is dominated by deformation to be distributed over a wide area (Koyi et
the convergence between the Arabian and Eurasian al., 2000). To the SE of Zagros, the Gowk fault separates
plates. The convergence is trending N to NNE at the Zagros collision zone from the Lut block with
velocity ranges from 23–25 mm/yr as deduced from relatively low levels of seismic activity (Berberian et al.,
GPS measurements (e.g., McClusky et al., 2003; 2000). Before the Bam earthquake, the Gowk fault was
Vernant et al., 2004) to 35 mm/yr according to the considered the only seismically active fault in the study
NUVEL-1 model (DeMets et al., 1990). To the west, the area (Ambraseys and Melville, 1982; Walker and
northwest-trending Zagros fold and thrust belt, which is Jackson, 2002). Five earthquakes of Mw = 5.4–7.1
an active continental collision zone, accommodates have occurred on this fault since 1981 (Berberian et
about 10 mm/yr of NNE-trending shortening (Alavi, al., 2001; Walker and Jackson, 2002), but all have been
1994; Talebian and Jackson, 2002; Tatar et al., 2002; more than 100 km from Bam. The Gowk fault to the
Blanc et al., 2003); also, in several areas further north, south, adjacent to the Bam fault, dies out in the Jebal
the crust is forced to accommodate the convergence by Barez Mountains, which themselves merge with the
shortening (Vernant et al., 2004). To the east, this active volcanic arc north of the Makran subduction zone
relative motion is accommodated by the east-trending (Walker and Jackson, 2002). The focal mechanisms of

Fig. 1. Simplified tectonic map around the study area. The arrows show Arabian plate motion relative to Eurasia. Convergence velocities are indicated
after the NUVEL-1 model (DeMets et al., 1990) and GPS studies (Tatar et al., 2002; Vernant et al., 2004). The strike-slip motion on the Gowk fault
comes from the tectonic work (Walker and Jackson, 2002). Three active volcanoes—Bazman, Taftan and Soltan—are associated with the Makran
subduction zone (Jakob and Quittmeyer, 1979). The location of the 2003 Bam earthquake and its focal mechanism (Yamanaka, 2003) are also shown.
272 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

the Gowk fault earthquakes (Berberian et al., 2001) 3. Data and method
suggest that transpressional deformation is active within
the study area. The Gowk and Bam faults are both of the The aftershock activity of the 2003 Bam earthquake
right-lateral strike-slip-type, transmitting the convergent was monitored by a seismic network consisting of nine
movement accommodated north of the Iranian plateau to temporal stations (Fig. 2). Each station was equipped
the Makran and accommodating the difference in with a highly sensitive three-component velocity-type
motion due to the transition between the Zagros seismometer (LE-3D, Lennartz Electronics) with a
collision and Makran subduction by transpressional natural frequency of 1Hz, and a GPS timing system
tectonics (see Regard et al., 2005 and references (Suzuki et al., 2004). The waveform data were
therein). continuously recorded at a sampling rate of 100Hz by

Fig. 2. (A) Hypocenter distribution (red dots) of the aftershocks located by 1-D velocity model (after Nakamura et al., 2005). (B) Hypocenter
distribution (red dots) relocated by using our 3-D velocity model. The lines AB, CD and EF correspond to the profiles of vertical cross-sections in Fig.
9. A NASA satellite map (http://earthobservatory.nasa.gov/) is shown. The green triangles denote stations of the temporal seismic network. Station 5
in Arg-e-Bam is also marked by a circle. The black dashed line indicates the traced line of the Bam fault inferred from the geological map supplied by
the National Geoscience Database of Iran (http://www.ngdir.ir/). Projections of the aftershock distributions on north–south (a) and east–west (c)
profiles are also shown.
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 273

computed 20 runs with different thicknesses of divided


layers in the initial model and finally obtained one of the
optimal 1-D models, as shown in Table 1, by linear
averaging the inverted velocity values of the adjacent
layers over 5km thicknesses. The selection of 5km
thickness in Table 1 is suitable for our 3-D model
parameterization, in which we adopted a grid spacing of
3–5km. We also tried an additional inversion with 5km
thick layers only by using P and S arrival times with rms
residuals of less than 0.1s of 308 aftershocks recorded
by all nine stations. The results showed that the
difference in velocity between the additional inversion
and Table 1 is less than 4%. We conclusively infer that
the 1-D model in Table 1 is appropriate for the initial
model in our 3-D tomography procedure. Using the
Wadati method (1933), we calculated a mean VP / VS
value through a polynomial regression on the differ-
Fig. 3. Wadati diagram for the aftershocks recorded by station 5 in Arg-
ences between S- and P-wave arrivals at station 5
e-Bam. Difference between P- and S-wave arrivals used for the
computation of VP / VS through a polynomial regression (gray line). (marked with a circle on Fig. 2-b). The seismograph at
station 5 was installed on an outcrop of volcanic hard
rock inside the Arg-e-Bam above the source fault.
a 16-bit data-logger (LS-8000SH, Hakusan). In general, Accurate S-wave arrival times were identified on the
the arrival time of the P phase could be identified to horizontal component of this station. Fig. 4 shows two
within about 0.05 s (sampling rate 100 Hz), whereas the examples of three-component seismograms: a very
estimation of the S-phase arrival was slightly less shallow aftershock and a relatively deep aftershock.
accurate. Using a one-dimensional (1-D) velocity model The polynomial regression performed in Fig. 3 gives
(Fig. 2A), we selected a useful data set of 2396
VP
aftershocks among the 2789 aftershocks that Nakamura ¼ 1:737−0:0489DTp
et al. (2005) accurately located by the double difference VS
method (Waldhauser and Ellsworth, 2000). All 2396 where ΔTp is a proxy for depth. The values VP / VS of
events were recorded by at least seven stations and by the 1-D model in Table 1 are consistent with this
the root mean square (rms) residual of P- and S-arrival polynomial regression. The station corrections of this 1-
times < 0.3s. The magnitudes of the aftershocks ranged D model are shown in Table 2 and Fig. 5. They vary
from M 0.1 to 3.1. from − 0.169 to 0.179 s for P waves and − 0.051 to
First, we constructed an initial model of 1-D velocity 0.488 s for S waves, correlating generally with the
structure for the 3-D tomographic inversion, as follows. surface geology. High positive values of more than
The initial parameters of the hypocenters were deter- 0.045 s for P waves and 0.243 s for S waves are related to
mined by HYPOMH software (Hirata and Matsu'ura, the stations of the southern side on Quaternary
1987) by use of the IASPEI standard velocity model. sediments, while low values of less than − 0.118 s for
After that we started searching for the best 1-D velocity P waves and 0.058 s for S waves are related to the
model by using the inversion program of Sato (1979). In northern stations on rocky sites.
each run, the program prescribes the 1-D model with the
minimum rms of residuals by working simultaneously
with 10 random initial models and an a priori model. Table 1
The initial models were constructed by using the Initial 1-D model of the velocity structure in Bam area used for 3-D
Gaussian random numbers around the a priori model travel time inversion
as the mean with a standard deviation. For the a priori Depth to bottom of VP Variance VS Variance VP/VS
model, we used the IASPEI1991 table because of the layer (km) (km/s) (km/s) (km/s) (km/s)
absence of reliable seismological information around 5 5.577 0.011 3.217 0.017 1.733
the Bam area. The standard deviations were within 10 5.830 0.030 3.464 0.018 1.683
0.25 km/s for P waves and 0.15km/s for S waves, 15 5.929 0.068 3.512 0.024 1.688
20 6.032 0.046 3.509 0.018 1.719
defined by plotting a Wadati diagram (Fig. 3). We
274 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

Fig. 4. Examples of three-component seismograms of a shallow aftershock (A) and a relatively deep aftershock (B) recorded at station 5 in Arg-e-Bam.

Secondly, we researched the 3-D velocity structures 19490 P- and 19015 S-wave arrival times, were used in
of P and S waves by using the tomography method this study. The unknown parameters are the hypocenters
developed by Zhao et al. (1992). We adopted the of aftershocks and velocities at the grid nodes, both of
pseudo-bending method (Um and Thurber, 1987) for ray which are determined in an iterative inversion process.
tracing and a conjugate gradient algorithm (Paige and The velocity at any point in the model is calculated by
Saunders, 1982) to invert the large and sparse system of linearly interpolating the velocities at the eight grid
observation equations that relate observed travel times nodes surrounding that point. As the dominant frequen-
to unknown parameters. A 3-D grid net was set up in the cy is 8–10Hz for P waves and 5–8 Hz for S waves, the
model space of the study area to express velocity corresponding Fresnel zones do not exceed 0.8km. This
structure. A total of 38505 phase data, consisting of is much smaller than the grid spacing of about 3–5km

Table 2
The data of seismic stations including station corrections obtained from 1-D travel time inversion
Station Longitude (°E) Latitude (°N) Elevation (m) Surface geology P-wave corr. (s) S-wave corr. (s)
1 58.2771 29.1929 1280 Well-bedded ash-flow tuffs with subordinate −0.149 − 0.051
2 58.2893 29.0615 1202 Quaternary sediments 0.121 0.386
3 58.2864 28.9592 1235 Quaternary sediments 0.128 0.410
4 58.3969 29.1607 1080 Well-bedded ash-flow tuffs with subordinate −0.155 0.015
5 58.3690 29.1160 1071 Volcanic rock (andesite) −0.162 0.058
6 58.3949 29.0084 1015 Quaternary sediments 0.045 0.279
7 58.4614 29.0607 952 Quaternary sediments 0.109 0.243
8 58.4731 29.1261 1006 Well-bedded ash-flow tuffs with subordinate −0.118 − 0.011
9 58.4556 28.9466 905 Quaternary sediments 0.179 0.488
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 275

Fig. 5. Simplified geological map of the Bam earthquake area based on the 1 : 100,000 geological map prepared by the Geological Survey of Iran,
Sheet 7648-Bam (1993). The grey triangles denote stations of the temporal seismic network. Crosses and circles on the stations show the positive and
negative station delays for the 1-D P-wave velocity model, respectively. The Bam seismic (accelerograph) station is marked by a grey diamond.

adopted in this study, and hence would not affect the study area. As the resolution naturally depends on the
tomographic imaging. After the VP and VS images are ray coverage, the denser regions are expected to have
determined from travel time inversion, we obtain the higher resolution. We chose grid nodes where more than
Poisson's ratio (σ) distribution by using the following 100 rays of P waves and more than 100 rays of S waves
relation.
 2
VP
2−
VS
r¼  2 :
VP
2−2
VS

4. Tomography results and resolution

Before describing the results of the 3-D tomographic


inversion, we first show the results of the checkerboard
resolution test (CRT). This test was conducted to assess
the ability of the data and the method to recover existing
velocity anomalies within the model. To make a
checkerboard, we assigned positive and negative
velocity perturbations with a 5% anomaly alternately
to all the 3-D grid nodes. We set up a 3-D grid in the
study area of 58.25E–58.50E and 28.9N–29.2N with a
grid spacing of 0.05° (about 5 km) in the horizontal
direction. Five layers of grid nodes are set up, one at
each of five depths: 0, 3, 6, 9 and 14km. Fig. 6 shows
the grid net distribution adopted in this inversion. We
obtained the results of the CRT at these five represen-
Fig. 6. Map view and cross-sectional view of the grid net adopted in
tative layers for both VP and VS structures, as shown in the 3-D inversion. Grid spacing in the horizontal direction is 0.05°
Fig. 7. The resolution test is generally good for the (∼5km). Five layers of grid nodes are set up at 0, 3, 6, 9 and 14km
layers, and synthetic anomalies are well recovered in the depths.
276 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

Fig. 7. Results of checkerboard resolution tests (CRT) for P-wave and S-wave velocity structures. The depth of each layer is shown at the bottom of
each map. Open and solid circles denote low and high velocities, respectively.

passed through. The average hit counts were 3222 and 5% anomalies were well reconstructed. We were
3140 for P and S rays, respectively. The starting velocity therefore able to use VP / VS = 1.73 to verify that the
model was the inverted 1-D model (Table 1) for the result is not affected by the grid configuration. After
P-wave velocity structure. The CRT showed that getting the results of CRT, we solved 2369 × 4
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 277

hypocenter parameters and 120 velocity parameters of Fig. 8 shows a map view of VP and VS and Poisson's
VP and VS at the grid nodes. Conclusively, the rms of ratio distributions at each depth layer. Significant
arrival time residuals was reduced by 17% and 12% for variations of up to 5% for VP and VS are revealed in
P waves and S waves, respectively. The final rms is the Bam area. The areas with the greater amplitude of
0.042 s for the P-wave data and 0.065 s for the S-wave the velocity anomalies are generally distributed in the
data, indicating the high quality of the arrival time data. first two layers, which are at depths of 0 and 3 km. At
Ray coverage and residual reduction can give a sense of these depths, the velocity patterns of P and S waves are
the reliability of the inversion results. However, it quite different from each other, thereby causing a high
should be noted that there are apparent limitations of amplitude of Poisson's ratio anomalies. At a depth of
seismic tomography (Zhao and Kayal, 2000). The main 0 km, high VP anomalies are located in the central part,
practical limitation of seismic tomography is the lack of including the city of Bam and the village of Baravat, and
any robust estimates of reliability in the resulting 3-D to the northeast of the study area. The S wave exhibits
models. We have also conducted an inversion by low-velocity anomalies except beneath of Baravat. The
adopting the inverted 1-D model in Table 1 for starting main feature of the layer at 0km depth is a high σ. At a
the S-wave structure. The inversion results show that the depth of 3 km, low VP anomalies are seen in the main
overall pattern of velocity structure is stable and that the center of the area, while high VS anomalies are pro-
change in velocity anomalies is generally less than 1.0% minent in the area surrounding this center. At this depth,
between VP / VS = 1.73 and Table 1 for the starting S- the Poisson's ratio pattern of the layer is quite different
wave structure. The velocity perturbations in Figs. 8 and from that of the layer at 0 km, and low σ anomalies are
9 are expressed in values relative to the 1-D inverted P- distributed widely. In the map of the 6-km-deep layer, a
wave structure and VP / VS = 1.73. maximum 2% anomaly of high VP beneath Bam is

Fig. 8. Plane views of perturbations of P velocity (top), S velocity (middle), and Poisson's ratio (bottom) in the study area. The depth of each layer is
shown at the bottom. Slow velocity and high Poisson's ratio are shown by red; fast velocity and low Poisson's ratio are shown by blue. The
perturbation scales are shown on the right side.
278 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

Fig. 9. Vertical cross-sections of P-wave and S-wave velocity perturbations and Poisson's ratio along the lines AB, CD and EF shown in Fig. 2B-b.
Slow velocity and high Poisson's ratio are shown by red; fast velocity and low Poisson's ratio are shown by blue. Small black crosses denote the
relocated aftershocks within a 2km width along the profile.

surrounded by low VP anomalies. The VS map shows a and 0.25 km for N–S, E–W and depth, respectively. The
maximum 5% of high velocity anomaly in the southeast comparison of hypocenter distributions between Fig. 2A
part of the study area and a strongly low velocity and B is discussed in the next section.
anomaly in the northwest part. The Poisson's ratio map
shows a high σ anomaly beneath Bam and a low σ 5. Discussion
anomaly in the southeast. At a depth of 9 km, the
significant linear trends of low VS and high σ appear in a We first discuss the 3-D images in the vertical cross-
north–south direction and along the Arg-e-Bam fault, sections of VP, VS and σ along and perpendicular to the
which Nakamura et al. (2005) proposed as the source Bam aftershock alignment (Fig. 9). The north–south
fault of the Bam earthquake. The maps of VP, VS and cross-section AB shows a general shape including the
Poisson's ratio at a depth of 14 km reveal no special earthquake fault plane. The east–west cross-sections CD
anomalies exceeding 1%. By using our 3-D velocity and EF are perpendicular to the fault plane at 29.10N
results, we relocated aftershocks and showed their through Bam and at 29.05N south of the city,
hypocenter distribution in Fig. 2B in comparison with respectively. Those images present large variations in
the one located by the 1-D velocity model (Fig. 2A, seismic velocity and Poisson's ratio. The cross-section
Nakamura et al. (2005)). The average location errors of CD along 29.10N in Fig. 9 shows a surface layer of high
the 3-D velocity results are estimated to be 0.10, 0.11 σ (0.28–0.30) down to a depth of about 1 km. This may
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 279

indicate a sediment layer of poorly consolidated fluids decreases the strength of rock through such
materials with a lot of water in the northern area of mechanisms as “stress corrosion cracking” (Atkinson
Bam, which is an oasis city. It is plausible that a well- and Meredith, 1987). Fluids can also weaken a rock and
developed flood plain associated with the Posht-rud enhance creep rates and slow deformation through a
River (Fig. 5) made such water-rich sediment. On the range of mechanisms (Etheridge et al., 1984; Tullis et
other hand, there is no such surface layer of high al., 1996). These would have enhanced stress concen-
Poisson's ratio in the southern area of Bam along tration in the seismogenic layer, leading to mechanical
29.05N, as shown in the cross-section EF. We avoid failure and thus contributing to the nucleation of the
discussing the cause of low VS and high σ in the shallow Bam earthquake, as we discuss latter. However, we have
layer south of 29.0N, because of the poor resolution by no exact explanation about the origin of the crustal
CRT in Fig. 7. The most prominent anomaly is a thick fluids in this region. We presented two causes, basalt
layer of low σ (0.20–0.24) at the depth range of about and fluids, of the high σ in the depth range of 6–9km,
2–6km, as is especially apparent in the cross-section EF. but have no definitive suggestions to make about them.
A teleseismic analysis (Yamanaka, 2003) indicated that Other forms of detailed prospecting, such as electro-
a large slip existed in a shallow part of the fault plane. magnetic prospecting, are expected, especially for
Using radar data, Wang et al. (2004) suggested that the researching fluids in the deep layers under the Bam area.
maximum slip occurred at a depth of about 3 km. And Secondly, we compare the distribution of aftershocks
Fialko et al. (2005) indicated that most of the seismic shown in Fig. 2B, which were located by the 3-D
moment was released at a depth of 4–5 km. Those depth velocity model, with that in Fig. 2A, which were
ranges correspond to approximately the central depth of obtained by the 1-D model. In both distributions, the
this layer with low VP (5.35–5.77 km/s), high VS (3.25– trends are in accord with the strike and dip angles of the
3.40 km/s) and low σ (0.20–0.24). The obtained VP focal mechanism (strike 175°, dip 85° and slip 153°) of
corresponds to the experimental value of 5.533 km/s for the main shock (Yamanaka, 2003). In addition, both
andesite under 200 MPa in Table 2 of Christensen epicenter distributions extend for about 20 km in the
(1996). But the obtained VS is slightly higher than the strike direction. However, the main linear distribution of
experimental value of 3.034 km/s for the same andesite the 3-D model in Fig. 2B shows a shift of about 1 km to
under 200 MPa. Referring to the experimental study of the west in comparison with that of 1-D in Fig. 2A. This
Christensen (1996), we would suggest that the observed suggests that the source fault of the main shock is about
lower σ indicates rock with higher SiO2 content and 4.5 km on average to the west of the geological Bam
greater brittleness. If the outcrops of andesite, trachyan- fault. Fig. 10 shows the same seismic cross-section, EF,
desite and dacite in the Bam area (Fig. 5) contain much as in Fig. 9, in contrast with the location of the Bam fault
quartz, this high quartz content could explain the on the ground surface (inverted arrow). We also show
obtained seismic velocities and Poisson's ratio. How- the location of minor surface ruptures with en-echelon
ever, we need more detailed petrological and petrophy- patterns in the desert presented in Fig. 5b of Fielding et
sical experimental studies of those rocks before we can al. (2005). The location of surface displacement was
reach a conclusion. nearly the same one modeled from satellite data (e.g.,
Fig. 9-c shows a high Poisson's ratio (0.27–0.28) in Talebian et al., 2004; Binet and Bollinger, 2005;
the depth range of about 6 km beneath Bam (29.11N, Fielding et al., 2005). In Fig. 10, the pattern of
58.35E) to about 9km south of there (29.02N, 58.35E) aftershocks deeper than 5km is complex. But the main
in the profile AB. The inferred increase in Poisson's linear cluster of aftershocks shallower than 5 km clearly
ratio may be associated with a change from a SiO2-rich faces just upward from the surface ruptures and nearly
rock to a more mafic rock. Among the mafic rocks, the touches them, coming within 1 km. Of course we have to
obtained values of VP (5.89–5.94 km/s) and VS (3.24– know that the average location errors of those
3.33 km/s) may be related to basalt under 200 MPa, the hypocenters that are shallower than 3 km are estimated
velocities of which are 5.914 km/s for P waves and to be 0.10, 0.11 and 0.32km for N–S, E–W and depth,
3.217 km/s for S waves, respectively, in Table 2 of respectively. We therefore propose a simple schematic
Christensen (1996). On the other hand, we suggest model of the central fault plane of the Bam earthquake in
another possible cause of the increase of Poisson's ratio: Fig. 10. As a matter of fact, the deeper part of the fault
the existence of fluids in the crust. Generally, fluid plane is thought to be more complex. If we assume the
saturation leads to an increase of Poisson's ratio (Ito et existence of a small second source fault (Funning et al.
al., 1979). Fluids can alter the rheology of rocks from (2005)) connecting to the Bam fault escarpment, we
brittle to ductile behavior. The chemical influence of could show that by the thick broken line labeled “2” in
280 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

nearly corresponds to the thick layer with high VS and


low σ, as shown in Fig. 9. On the condition of the same
density, the high VS means higher rigidity in this layer in
comparison with the surroundings. In addition, referring
to the teleseismic analyses of Yamanaka (2003) and
Yagi (2003), we suggest that the start of the higher slip is
near the bottom of the asperity. This means that the
rupture of the Bam earthquake started in or near the
deeper layer of high σ, even though we cannot show its
exact point or the precise hypocenter of the main shock
in Fig. 11. This allows us to assume that the nucleation
of the Bam earthquake was created in this deeper layer
filled with fluids. We surmise that the rupture of the
main shock started in or near the high σ layer of 6–9km
in depth and then propagated with a large slip mainly in
the rigid rocks at the depth range of 2–6 km. This
Fig. 10. Interpretative cross-section EF in Fig. 9 showing the areas of rupture with a large slip must have generated the very
low Poisson's ratio (Lσ: 0.20–0.24) and high Poisson's ratio (Hσ: strong motions on the surface and cause intense damage
0.27–0.28) by thin broken lines and the distribution of aftershocks in and around Bam. By using the acceleration data
within a 2km width along the profile (black crosses). A schematic observed at the Bam station and other stations (BHRC),
simple model of the central fault plane of the Bam earthquake is shown
by a thick solid line marked with the number 1. A possible branching
Miyake et al. (2004) suggested that the extremely strong
segment of the fault plane is also shown by a broken line marked with motions were localized and proposed the shallow
the number 2. The locations of minor surface ruptures (Fielding et al., asperities existed on a fresh fault rather than the Bam
2005) and the Bam fault on the ground are indicated by inverted fault. We cannot judge between “fresh” and “not fresh”.
arrows marked with SR and BF, respectively. The locations of cross- But our shallow asperity model in the rigid rocks can be
sections AB (shown in Fig. 9) is also indicated.

Fig. 10. On the other hand, we suggest that the northern


part of the fault under Bam, including Arg-e-Bam, is not
a single plane but branching planes (Nakamura et al.,
2005), because the whole pattern of aftershocks in the
cross-section CD in Fig. 9 is very complex.
Fig. 11 shows the same seismic cross-section AB as
in Fig. 9, including the areas of low and high σ. Most
aftershocks are distributed between 0 and 14 km in depth
and became shallower than that in Fig. 2A. The seismic
gap in the central part of the cross-section can be
distinguished at the depth range of 2–7 km, as shown by
the dotted circle. This may correspond to the higher slip
region (asperity) of 80cm to 1m proposed by Yamanaka
(2003) as a result of teleseismic analysis. In Fig. 11 we
also show the area with a slip larger than 2 m of the main
fault by the satellite data analysis of Funning et al.
(2005). This figure suggests that our seismic gap area
presents not a perfect but rather a good coincidence with
the large slip area of Funning et al. (2005). We can say
that the asperity of the Bam earthquake was very Fig. 11. Interpretative cross-section AB in Fig. 9 showing the areas of
shallow, nearly in the depth range of 2–7km. This low Poisson's ratio (Lσ: 0.20–0.24) and high Poisson's ratio (Hσ:
shallow asperity must be one of the reasons why the 0.27–0.28) by thin broken lines. The dotted circle shows the seismic
gap in the central part of the cross-section. The area with a slip larger
damage was unexpectedly large given the earthquake's than 2m of the main shock (Funning et al., 2005) is shown by the thick
magnitude. We also suggest that the asperity inferred broken line. The locations of cross-sections AB and CD (shown in Fig.
from the seismic gap in the depth range of 2 to 7km 9) are indicated by inverted arrows.
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 281

supported by the localized nature of the extremely References


strong motions.
Ahmadizadeh, M., Shakib, H., 2004. On the December 26, 2003,
southeastern Iran earthquake in Bam region. Engineering Struc-
6. Conclusions
tures, vol. 26, pp. 1055–1070. doi:10.1016/j.engstruct.03.006.
Alavi, M., 1994. Tectonics of the Zagros orogenic belt of Iran: new
Three-dimensional VP, VS and Poisson's ratio in the data and interpretations. Tectonophysics 229, 211–238.
epicenter area of the 2003 Bam earthquake have been Ambraseys, N.N., Melville, C.P., 1982. A History of Persian
determined by using a large number of high-quality Earthquakes. Cambridge Earth Science Series, Cambridge Uni-
versity Press, London. 219 pp.
arrival times from the aftershocks. A precise aftershock
Atkinson, B.K., Meredith, P.G., 1987. The theory of subcritical crack
distribution, relocated by using the 3-D structure, clearly growth with applications to minerals and rocks. In: Atkinson, B.
delineates the fault plane about 4.5km west of the K. (Ed.), Fracture Mechanics of Rock. Academic Press, London,
known Bam fault. The aftershock distribution in the pp. 111–166.
along-strike cross-section illuminates a rectangular fault Berberian, M., 1976. Quaternary faults in Iran. In: Berberian, M. (Ed.),
Contribution to the Seismotectonics of Iran (Part II). Geological
area of about 20 km in horizontal length and 14 km in
Survey of Iran, pp. 187–258. Report No. 39.
deep width from near the earth's surface. The fault plane Berberian, M., Jackson, J.A., Qorashi, M., Talebian, M., Khatib, M.,
dips westward steeply and strikes nearly north–south. Priestley, K., 2000. The 1994 Sefidabeh earthquakes in eastern
The main shallow cluster of aftershocks south of Bam Iran: blind thrusting and bedding-plane slip on a growing anticline,
city is distributed just under the small ruptures found on and active tectonics of the Sistan suture zone. Geophys. J. Int. 142,
283–299.
the ground surface. The 3-D structures of the seismo-
Berberian, M., Jackson, J.A., Fielding, E.J., Parsons, B.E., Priestley, K.,
genic region are well resolved to a depth of 14 km. A Qorashi, M., Talebian, M., Walker, R., Wright, T.J., Baker, C.,
thick layer of high VS and low σ anomalies (0.20–0.24) 2001. The 1998 March 14 Fandoqa earthquake (Mw 6.6) in Kerman
is imaged at about 2–6 km depth. Low σ may suggest province, southeast Iran: re-rupture of the 1981 Sirch earthquake
that the rocks have high SiO2 content. A high σ (0.27– fault, triggering of slip on adjacent thrusts and the active tectonics
of the Gowk fault zone. Geophys. J. Int. 146, 371–398.
0.28) zone is clearly imaged in the depth range from Binet, R., Bollinger, L., 2005. Horizontal coseismic deformation of the 2003
about 6km beneath Bam (29.11N, 58.35E) to about Bam (Iran) earthquake measured from SPOT-5 THR satellite imagery.
9 km south of the city (about 29.02N, 58.35E). This zone Geophys. Res. Lett. 32, L02307. doi:10.1029/2004GL021897.
may suggest a change from SiO2-rich rock to a more Blanc, E.J.-P., Allen, M.B., Inger, S., Hassani, H., 2003. Structural
mafic rock, or it may suggest the existence of fluids. The styles in the Zagros simple folded zone, Iran. J. Geol. Soc. London
160, 401–412.
main seismic gap of aftershock distribution at the depth
Byrne, D.E., Sykes, L.R., Davis, D.M., 1992. Great thrust earthquakes
range of 2 to 7km appeared in nearly good coincidence and aseismic slip along the plate boundary of the Makran
with the large slip zone in the shallow thick layer of high subduction zone. J. Geophys. Res. 97, 449–478.
VS and low σ. This high VS and low σ may appear to Chiarabba, C., Selvaggi, G., 1997. Structural control on fault
indicate high rigidity and brittleness in comparison with geometry: example of the Grevena Ms 6.6, normal faulting
earthquake. J. Geophys. Res. 102, 22445–22457.
the surroundings. We therefore conclude that the large
Chiarabba, C., Amato, A., Meghraoui, M., 1997. Tomographic images
slip propagating mainly in the shallow rigid layer in and of the El Asnam fault zone and the evolution of a seismogenic
south of Bam is one of the main reasons why the Bam thrust related fold. J. Geophys. Res. 102, 24485–24498.
area suffered heavy damage. Christensen, N., 1996. Poissons's ratio and crustal seismology.
J. Geophys. Res. 101, 3139–3156.
DeMets, C., Gordon, R.G., Argus, D.F., Stein, S., 1990. Current plate
Acknowledgments
motions. Geophys. J. Int. 101, 425–478.
Eberhart-Phillips, D., 1990. Three-dimensional P and S velocity
We gratefully acknowledge Dr. T. Matsushima, Dr. structure in the Coalinga region, California. J. Geophys. Res. 95,
T. Ito, Dr. S.K. Hosseini, A.J. Gandomi, and M. Maleki 15343–15363.
for their help with fieldwork and data analysis as well Eberhart-Phillips, D., 1993. Local earthquake tomography: earthquake
source regions. In: Iyer, H.M., Hirahara, K. (Eds.), Seismic
as for our constructive discussions with them. We thank
Tomography: Theory and Practice. Chapman and Hall, London,
Professor Dr. Dapeng Zhao for the tomography code pp. 614–643.
and Professor Dr. Tamao Sato for the one-dimensional Etheridge, M.A., Wall, V.J., Cox, S.F., Vernon, R.H., 1984. High fluid
velocity inversion code. Some figures in this paper pressure during regional metamorphism and deformation: implica-
were made using Generic Mapping Tools (GMT) tions for mass transport and deformation mechanisms. J. Geophys.
Res. 89, 4344–4358.
software written by Wessel and Smith (1998). The
Fialko, Y., Sandwell, D., Simons, M., Rosen, P., 2005. Three-
manuscript was greatly improved by comments and dimensional deformation caused by the Bam, Iran, earthquake and
suggestions of Professor Mike Sandiford and two the origin of shallow slip deficit. Nature 435, 295–299.
anonymous reviewers. doi:10.1038/nature03425.
282 H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283

Fielding, E., Talebian, M., Rosen, P., Nazari, H., Jackson, J.A., Suzuki, S., Fatemi Aghda, S.M., Nakamura, T., Matsushima, T., Ito,
Ghorashi, M., Walker, R., 2005. Surface ruptures and building Y., Sadeghi, H., Maleki, M., Gandomi, A.J., Hosseini, S.K., 2004.
damage of the 2003 Bam, Iran, earthquake mapped by satellite Temporal seismic observation and preliminary hypocenter deter-
synthetic aperture radar interferometric correlation. J. Geophys. mination of aftershocks of the 2003 Bam earthquake, southeastern
Res. 110, B03302. doi:10.1029/2004JB003299. Iran. Bull. Earthq. Res. Inst. Univ. Tokyo 79, 37–45.
Funning, G.J., Parsons, B., Wright, T.J., Jackson, J.A., Fielding, E.J., Talebian, M., Jackson, J.A., 2002. Offset on the main recent fault
2005. Surface displacements and source parameters of the 2003 of NW Iran and implications for the late Cenozoic tectonics
Bam (Iran) earthquake from Envisat advanced synthetic aperture of the Arabia–Eurasia collision zone. Geophys. J. Int. 150,
radar imagery. J. Geophys. Res. 110, B09406. doi:10.1029/ 422–439.
2004JB003338. Talebian, M., Fielding, E.J., Funning, G.J., Ghorashi, M., Jackson, J.,
Hessami, K., Jamali, F., Tabassi, H., 2003. Major Active Faults of Iran, Nazari, H., Parsons, B., Priestley, K., Rosen, P.A., Walker, R.,
Edition 2003. International Institute of Earthquake Engineering Wright, T.J., 2004. The 2003 Bam (Iran) earthquake: rupture of a
and Seismology. 1 Sheet. blind strike-slip fault. Geophys. Res. Lett. 31, L11611.
Hirata, N., Matsu'ura, M., 1987. Maximum-likelihood estimation of doi:10.1029/2004GL020058.
hypocenter with origin time eliminated using nonlinear inversion Tatar, M., Hatzfeld, D., Martinod, J., Walpersdorf, A., Ghafori-
technique. Phys. Earth Planet. Inter. 47, 50–61. Ashtiany, M., Chéry, J., 2002. The present-day deformation of the
Ito, H., DeVilbiss, J., Nur, A., 1979. Compressional and shear waves in central Zagros from GPS measurements. Geophys. Res. Lett. 29.
saturated rock during water–steam transition. J. Geophys. Res. 84, doi:10.1029/2002GL015427.
4731–4735. Tatar, M., Hatzfeld, D., Moradi, A.S., Paul, A., 2005. The 2003
Jakob, K.H., Quittmeyer, R.L., 1979. The Makran region off Pakistan December 26 Bam earthquake (Iran), Mw 6.6, aftershock
and Iran: trench–arc system with active plate subduction. In: sequence. Geophys. J. Int. 163, 90–105.
Farah, A., DeJong, K.A. (Eds.), Geodynamics of Pakistan. Tullis, J., Yund, R., Farver, J., 1996. Deformation-enhanced fluid
Geological Survey of Pakistan, Quetta, pp. 305–317. distribution in feldspar aggregates and implications for ductile
Kopp, C., Fruehn, J., Flueh, E.R., Reichert, C., Kukowski, N., Bialas, shear zones. Geology 24, 63–66.
J., Klaeschen, D., 2000. Structure of the Makran subduction zone Um, J., Thurber, C., 1987. A fast algorithm for two-point seismic ray
from wide-angle and reflection seismic data. Tectonophysics 329, tracing. Bull. Seismol. Soc. Am. 77, 972–986.
171–191. Vernant, Ph., Nilforoushan, F., Hatzfeld, D., Abbasi, M.R., Vigny, C.,
Koyi, H.A., Hessami, K., Teixell, A., 2000. Epicenter distribution and Masson, F., Nankali, H., Martinod, J., Ashtiani, A., Bayer, R.,
magnitude of earthquakes in fold-thrust belts: insights from Tavakoli, F., Chéry, J., 2004. Contemporary crustal deformation
sandbox models. Geophys. Res. Lett. 27, 273–276. and plate kinematics in Middle East constrained by GPS
McCall, G.J.H., 1997. The geotectonic history of the Makran and measurements in Iran and northern Oman. Geophys. J. Int. 157,
adjacent areas of southern Iran. J. Asian Sci. 15, 517–531. 381–398.
McClusky, S., Reilinger, R., Mahmoud, S., Ben Sari, D., Tealeb, A., Wadati, K., 1933. On the travel time of earthquake waves. Geophys.
2003. GPS constraints on Africa (Nubia) and Arabia plate motions. Mag. 7, 101–111.
Geophys. J. Int. 155, 126–138. Waldhauser, F., Ellsworth, W.L., 2000. A double-difference earth-
Miyake, H., Koketsu, K., Mostafaei, H., 2004. Rupture process of the quake location algorithm: method and application to the northern
2003 Bam, Iran, earthquake: did shallow asperities on a fresh Hayward fault, California. Bull. Seismol. Soc. Am. 90,
fault cause extreme ground motions? EOS Trans. Am. Geophys. 1353–1368.
Union 85. Walker, R., Jackson, J.A., 2002. Offset and evolution of the Gowk
Nakamura, T., Suzuki, S., Sadeghi, H., Fatemi Aghda, S.M., fault, S.E. Iran: a major intra-continental strike-slip system. J.
Matsushima, T., Ito, Y., Hosseini, S.K., Gandomi, A.J., Maleki, Struct. Geol. 24, 1677–1698.
M., 2005. Source fault structure of the 2003 Bam earthquake, Wang, R., Xia, Y., Grosser, H., Wetzel, H.U., Zschau, J., Kaufmann,
southeastern Iran, inferred from the aftershock distribution and its H., 2004. The 2003 Bam (SE Iran) earthquake: precise source
relation to the heavily damaged area: existence of the Arg-e-Bam parameters from satellite radar interferometry. Geophys. J. Int. 159,
fault proposed. Geophys. Res. Lett. 32, L09308. doi:10.1029/ 917–922.
2005GL022631. Wessel, P., Smith, W.H.F., 1998. New improved version of
Paige, C.C., Saunders, M.A., 1982. LSQR: an algorithm for sparse Generic Mapping Tools released. EOS Trans. Am. Geophys.
linear equations and sparse least squares. ACM Trans. Math. Union 79, 579.
Softw. 8, 43–71. Yagi, Y., 2003. Preliminary results of rupture process for 2003 De-
Regard, V., Bellier, O., Thomas, J.-C., Bourlès, D., Bonnet, S., cember 26 southeastern IRAN, Earthquake. International Institute
Abbassi, M.R., Braucher, R., Mercier, J., Shabanian, E., of Seismology and Earthquake Engineering. Available at http://
Soleymani, Sh., Feghhi, Kh., 2005. Cumulative right-lateral iisee.kenken.go.jp/staff/yagi/eq/Iran20031226/IRAN20031226.
fault slip rate across the Zagros–Makran transfer zone: role of htm.
the Minab–Zendan fault system in accommodating Arabia– Yamanaka, Y., 2003. Seismological Note: No. 145, Earthquake
Eurasia convergence in southeast Iran. Geophys. J. Int. 162, Information Center, Earthquake Research Institute, University of
177–203. Tokyo. Available at http://www.eri.u-tokyo.ac.jp/sanchu/Seismo_
Salah, M.K., Zhao, D., 2003. 3-D seismic structure of Kii Peninsula in Note/EIC_News/031226f.html.
southwest Japan: evidence for slab dehydration in the forearc. Zhao, D., Kayal, J.R., 2000. Impact of seismic tomography on Earth
Tectonophysics 364, 191–213. sciences. Curr. Sci. 79, 1208–1214.
Sato, T., 1979. Velocity structure of the crust beneath the northeastern Zhao, D., Hasegawa, A., Horiuchi, S., 1992. Tomographic imaging of
part of Honshu, Japan, as derived from local earthquake data. J. P and S wave velocity structure beneath northeastern Japan. J.
Phys. Earth 27, 239–253. Geophys. Res. 97, 19909–19928.
H. Sadeghi et al. / Tectonophysics 417 (2006) 269–283 283

Zhao, D., Kanamori, H., Negishi, H., Wiens, D., 1996. Tomography of Zhao, D., Tani, H., Mishra, O.P., 2004. Crustal heterogeneity in the
the source area of the 1995 Kobe earthquake: evidence for fluids at 2000 western Tottori earthquake region: effect of fluids from slab
the hypocenter? Science 274, 1891–1894. dehydration. Phys. Earth Planet. Inter. 145, 161–177.
Zhao, D., Mishra, O.P., Sanda, R., 2002. Influence of fluids and
magma on earthquakes: seismological evidence. Phys. Earth
Planet. Inter. 132, 249–267.

Вам также может понравиться